Traffic Sign Detection using Deep Learning

Computer Science and Engineering

For autonomous driving systems, classifying traffic signs is a crucial task. Traffic signs vary greatly in appearance depending on the nation, which makes it more difficult for classification systems to be successful. A larger collection of images should be used, or the classifier should be improved. Advanced Driver Assistance Systems (ADAS) and Intelligent Autonomous Vehicles (IV) are currently used to address the issue of traffic sign recognition. Due to the various and intricate situations, they are put in, it is a difficult real-world computer vision problem. Images are grouped into categories like highway signs, speed signs, danger signs, etc. after being categorised. b In this project, we propose to investigate the YOLO Architecture and its compatibility in order to solve this problem. The goal is to find and classify traffic signs in natural street scenes. The main challenge in this problem is recognising minute targets in a large and complex image background. Other object detection models, such as Fast RCNN and Faster R-CNN, has been used to solve this problem. The main disadvantage of such methods is their slowness - they are not real-time. Thus, the motivation for investigating YOLO for this task is speed - it is about 6 faster than faster R-CNN. In this paper, we also propose a novel The modified loss function for the YOLO model to improve its performance in traffic sign detection.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.